Advancing Multi-step Mathematical Reasoning in Large Language Models Through Multi-layered Self-reflection with Auto-prompting Chapter in Scopus